There has been reports of very noisy electret mics
https://electronics.stackexchange.c...plifying-and-filtering-circuit-output-so-much

I've got some electret mic from Aliexpress. They looked similar to these, e.g. the unconnected one in the middle.

these couple of days I ran some tests with a little DAQ/Oscilloscope i made with a stm32. the circuit is like such

VCC 5v +-----R 680 ohm -----+------- electret mic ------+ GND

Then center tap goes to the ADC input, originally I recorded the most significant 8 bits, converted them to wav files. the 'scope' output looks like this.
ok 20 mV peak to peak.

I listened to the wav record, and it turns out it is all noise, there is no audible sounds recorded.

So I thought maybe 8 bits is too coarse, I updated my adc firmware source codes to extract the full 12 bits adc samples.
Now i've got 1mV precision. So i recorded that converted to wav format, import that in Audacity https://www.audacityteam.org/ post processed it a little. The post processing involves normalization rebased signal average (around 2.63V) to zero (similar to putting a series capacitor), digitally amplifying that 10mv signals so that it occupies the full digital peak to peak amplitudes, low pass filter at 20khz (sample rate 44khz) remove aliasing.
after that processing it looks like this.
It is still wide band noise and nothing, i clapped a couple times near the mic which is rather loud, but that clap did not present itself in the audio record, all noise. No audible sounds. The wav recording is in rec1.zip attached.

 

From the photograph it looks like you need to turn the mic around. It is polarity sensitive.

 

I had a lot of trouble with these a couple of decades ago. The cheap ones have the gate lead of the JFET held to the element by nothing more than the case crimp. The better ones (after that we only used Panasonic) have it welded.
You can often quieten them down by tightening up the crimp, but first, check the polarity as @MrChips says, it also looks to me as though it’s back to front.

 

if i turn the mic around i'd not get 2.63v, I get some values can't remember but it is something like close to zero or that it saturates.
while if i followed the circuit presented in this specs
https://mikroshop.ch/pdf/MIKRO_XL.pdf
which is pretty much in line with the specs I got 2.63v out of 5v vcc, across the electret mic.

For now i've taken apart the circuits on the breadboard, i tried several from the same order about 10 pieces, they behaved the same.
i'd guess i'd try some other electrets from other manufacturers. the cheap ones, they 'looked like' electrets complete with a felt cover etc. but whether it works well is another thing. I almost want to make a preamp circuit for it with an op amp, but I decided it is a waste of time to do that given that measured down to millivolts i get noisy signals, amplifying it would likely simply amplify the noise.
the noisy signals may have come from other sources as this is patched onto a breadboard and measured, radio waves rf noise sources i'd guess can easily contribute those few millivolts noises.

 

In the data sheet they used Vs = 1.5V
You are using 5V.
Increase the value of the load resistor. Try 4.7kΩ for starters.

 

Have you tried the MEMS microphones?
I found them to be rather better than the old telephone inserts, not exactly studio quality, but pretty good.

 

i'd think MEMS microphones should work, but unlike 'cheap' electrets, MEMS probably requires a proper semiconductor fab to make them. While i'd guess for 'cheap' white labelled electrets, one takes a gamble and performance varies. It is hard to tell if that little aluminium can which looks like one in a spec is after all the same thing.

 

ok I finally got the electret mic bought on Aliexpress to register some sounds. I did that by knocking on the breadboard while recording. This is processed somewhat to remove the background high frequency noise. There are 5 knocks altogether.


There are some background sounds but that non of the background sounds can be heard in the recording. Only the knocks which creates rather strong electrical impulses register as sounds.
Hence, these 'cheap' electrets are not very sensitive really.
The wav file is in the zip file attached.

 

Another recording, this time i got some music recorded. This is done by placing the phone speaker against the mic, in contact. the record looks like this

it turns out this needs to be done at vcc 3.3v, dc voltages at the mic is 3.1v leaving just 0.2v as a range, R is 680 ohms , this means that the JFET in the electret is practically turned off. and that the speaker needs to be turned up at loud volume for this to be registered. lower speaker volume playing it and nothing is registered. it really isn't very sensitive.

i've not been able to reproduce this at 5v vcc, i'm not too sure why that's the case though. At 5v vcc, the operating voltages certainly looked better. and the range (judged by the noise range alone) is about double that at 3.3v.

the wav file is in the zip file attached.

 

It looks like it is working but not as sensitive as it should be. Did you try changing the load resistor as suggested? The doc you linked calls for 2.2K.

Bob

 

thanks i'd try changing the load resistor to see if it helped, maybe that'd make it work at 5v.
taking a break for now. At least that they worked, but i'd guess one should temper expectations for these white labelled electrets sold on ebay, aliexpress etc. they are likely less sensitive vs brand name ones. and though the web is littered with specs floating around that seem to match the *look* of the same electrets. I'd guess it is quite a gamble, lottery, if you happened to buy from a vendor who stock better electrets, chances are that they worked better than this. But i bought the 'cheap' ones as specs with the matching 'looks' seemed good and they seemed well made with that felt cover and all. performance is quite another thing from that 'look'

 

ok here is an update, i used a 4 k ohm load resistor instead, VCC remains at 3.3v
now with the 4 k ohm load resistor, dc voltages at the mic is 2.29v that is much better. this center tap goes to the adc, nothing sits between.
i ran the same recording, mobile phone speaker against mic, turn up volume about 70% full volume - that is pretty loud, audible near it about 2 feet
this time the 'scope' looks much better, that high band noise is still visible and from the graphs it is about 10-30mV pretty bad.
in this graph each tick mark is 1ms.
I tried putting a 470 uF capacitor between VCC (3.3v) and GND but it did not seem to eliminate the noise, this record is done with the 470 uF cap in place. From the graph alone, it would seem the peak to peak voltages at the mic can be about 0.1v (100 mV), that would have been enough for a 8 bits adc to detect it. Nevertheless, this record is done at full 12 bits, this also gives a better illustration of the noises riding on the actual sound waveforms.


In audacity the record looks like this. post processing is done, normalize rebase 2.8v to zero, amplify it (that 100 mV to PCM amplitudes) to occupy the pcm amplitudes and to low pass filter it at 20000 khz (anti-alias), results looks like this.


This time round that lousy music is due to my lousy recording firmware and capture implementation, lol.
The file is attached as an mp3 file in the zip attachment.

for those curious about where i got it, they are found with a search for electret as keyword
https://www.aliexpress.com/wholesale?catId=0&SearchText=electret
it is those "tin/aluminium cans with a black felt cover and at the GND contact, there are 3 'wiskers' on the pcb to show that it is GND"
i'm not too sure if they came from the same or few manufacturers.

 

when i tried repeating the same recording again by playing the song over again, mic against mobile phone speaker 70% volume.
this time there are other artifacts

on the 'scope' it looks like this


nothing moves in that high noise damaged portion between 15-40 secs. they sounded as 'clicking' noises mingled with music.
I think this is in line with what Ian0 mentioned, the FET contacts are just 'touching' the sensor surface rather than soldered.
That could account for such behaviours.

the damaged mp3 file in the zip attachment.

 

Why are you using a defective "cheap" microphone from ebay or AliExpress when a Name-Brand one is sold at Digikey for only $1.02US?
Get rid of the breadboard and make good soldered connections to a shielded audio cable.
Most electret mics draw 0.4mA (0.5mA max) and need at least 2V across the mic so use a 5.6k series resistor and a 5V supply.

 

You can get new electret microphones with a datasheet and the name of the manufacturer for about U.S. 36¢ each.
http://www.es.co.th/detail.asp?Prod=004400131

The stuff from Aliexpress and eBay and some discount hobby suppliers is often not what is advertised or are manufacturing seconds (rejects) and are fun to play with, but when you need results to count on, it is best to get something from a distributor with a good reputation.

 

thanks, it so happen I got a handful of them from Aliexpress on impulse. they have been left in the gadget pile for a while and while searching for something, pounced on it. Hence, i started testing them out, an experiment kind of.
for what is worth, i pounced on some datasheets on the web and they looked like these electrets, complete with that '3 whiskers' to denote GND on the pcb etc. The specs apparently specify something much better than what is observed in these experiments.

it is fairly apparent now that 'cheap' white labelled electrets some of which like these I got are *not very sensitive*, it can hardly pick up any background sounds and requires rather loud volume just to register the electrical impulses. i'd think it'd take like speaking near it like at a 4 inch distance and rather loudly. I'm still experimenting to figure out the nominal sensitivity. Background sounds that you would normally hear don't get registered as impulses, they are hardly visible on the 'scope' measured down to millivolts precision.

But yes those from digikey etc are quality assured.
A thought though about MEMS mics, MEMS mics are most likely the better mics as those are used in mobile phones.
MEMS mics likely has much better quality assurance as they are manufactured and quality controlled in the fabs and are inexpensive due to mass production for mobile phones

edit:
I posted some initial thinking about an electret mic amp here
https://forum.allaboutcircuits.com/threads/non-inverting-inverting-lm358.184756/post-1706185

 

since I've a bunch of them I tested a few (3) pieces randomly selected, it turns out all 3 electrets exhibits the same high noise damaged sections and it is apparently rather random. But I noted that they tend to coincide with sections of higher audible volume. recording background sounds draws a blank, apparently the background sounds in the low 10s of millivolts (say 0-15 mV) are kind of 'lost' in the high band noise. the high band noise could have origins other than the electret.
but those characteristic 'damaging' sections (the random click sounds), I'd think is a feature/defect of these electrets.

 

1) Your supply voltage is too low.
2) The resistor powering the mic is so low that it is shorting out the mic signal level.
3) An LM358 is never used as a mic preamp because it has too much noise, it has crossover distortion and has a poor high frequency response.

Cheap mics and other things are probably factory defects or fakes.